基于上下文信息增强和深度引导的单目3D目标检测

于家艺,吴秦

Monocular 3D object detection based on context information enhancement and depth guidance

Jiayi YU,Qin WU

表 1 不同目标检测模型在KITTI测试集上的单目3D目标检测精度对比

Tab.1 Comparison of monocular 3D object detection accuracy for different object detection models in KITTI test set %

模型 发表位置 $ {{\mathrm{AP}}_{{\mathrm{3D}}}}|{R_{40}} $ $ {{\mathrm{AP}}_{{\mathrm{BEV}}}}|{R_{40}} $ 简单 中等 困难 简单 中等 困难 CaDDN[21] CVPR21 19.17 13.41 11.46 27.94 18.91 17.19 Monodle[22] CVPR21 17.23 12.26 10.29 24.79 18.89 16.00 GrooMeD-NMS[23] CVPR21 18.10 12.32 9.65 26.19 18.27 14.05 MonoEF[24] CVPR21 21.29 13.87 11.71 29.03 19.70 17.26 MonoFlex[25] CVPR21 19.94 13.89 12.07 28.23 19.75 16.89 AutoShape[7] ICCV21 22.47 14.17 11.36 30.66 20.08 15.95 GUPNet[10] ICCV21 22.26 15.02 13.12 30.29 21.19 18.20 PCT[26] NeurIPS21 21.00 13.37 11.31 29.65 19.03 15.92 MonoGround[27] CVPR22 21.37 14.36 12.62 30.07 20.47 17.74 HomoLoss[28] CVPR22 21.75 14.94 13.07 29.60 20.68 17.81 MonoDTR[14] CVPR22 21.99 15.39 12.73 28.59 20.38 17.14 MonoJSG[29] CVPR22 24.69 16.14 13.64 32.59 21.26 18.18 DCD[9] ECCV22 23.81 15.90 13.21 32.55 21.50 18.25 DEVIANT[30] ECCV22 21.88 14.46 11.89 29.65 20.44 17.43 DID-M3D[17] ECCV22 24.40 16.29 13.75 32.95 22.76 19.83 SGM3D[31] RAL22 22.46 14.65 12.97 31.49 21.37 18.43 MonoCon[32] AAAI22 22.50 16.46 13.95 31.12 22.10 19.00 MonoRCNN++[33] WACV23 20.08 13.72 11.34 — — — MonoEdge[34] WACV23 21.08 14.47 12.73 28.80 20.35 17.57 本研究 — 26.74 16.67 14.33 34.73 22.84 19.52